February 21, 2007

SemFest is coming!

The annual Stanford Semantics Festival is coming, and I have two papers on the program (making up for last year, when I was on leave and didn't submit a paper).  It's a local tradition, and a very pleasant one: serious, but small, mixing students and faculty, with people from several departments.  Regularly on the last day of classes in Winter Quarter,

This year, both of my papers are connected to things I've been posting about here on Language Log.  Below is a slightly expanded version of one of them.  (My apologies for its being in abstract-speak rather than in more conversational prose.)  The other will follow in, as they say, due course.

Avoid vagueness?  The case of sentence-initial linking however
Arnold M. Zwicky & Douglas W. Kenter

When two items are very similar in meaning, but one (the Special alternative) is in some way more specific than the other (the General alternative), the general maxim Avoid Vagueness (AV) comes into play:

AV: Be specific; avoid vagueness.

The straightforward way to obey AV in the case of Special/General pairings would be to adhere to the guideline Just as Specific as Necessary (JASAN):

JASAN: Use Special when it is appropriate; otherwise, use General.  

If you follow JASAN, Special maintains its meaning, and General picks up some content by implicature.  An example: for the choice between the intensifier very and a more specific alternative like extremely, following JASAN keeps extremely towards the high end of the scale, while very continues to denote something up the scale, but now conveys that it's not at the high end.

Prescriptions about usage rarely suggest JASAN; instead, they routinely advise that AV be satisfied by avoiding General:

ALS: Avoid the Less Specific.

In particular, advice manuals routinely suggest avoiding very.  The consequence of following ALS is to move Special into the space General used to occupy, eventually bleaching it (and General as well).

In at least one case, some manuals actually advise AGAINST Special: from Strunk (1918) to recent works by Garner, we're told not to use sentence-initial linking however, as in:

The roads were impassible.  However, we at last succeeded in reaching camp.

Garner recommends using but instead.  (Call this Garner's Rule, GR, since Garner is its most energetic current expositor.  GR, of course, contradicts a widely touted but quite spurious "rule" No Initial Coordinators (NIC), barring sentence-initial coordinating conjunctions; see Zwicky 2006a,b.)  These advisers think of this however and but as equivalent in "meaning", and sense that however somehow weakens the effect of the clause that follows. (In what follows, references to however and but are to these two items as sentence-initial linking elements, unless otherwise noted.)

But however and but aren't equivalent: but is General, and however Special, as observed by Fraser (1998).  There are many circumstances (some of which we survey) where but is fine but however is at best odd -- in protests, for example:

A: It's bedtime.
B: But I haven't had a story yet. / *However, I haven't had a story yet.

In Schiffrin's (1987) terms, but marks a main unit in discourse organization, while however marks a subordinate unit (and so conveys more about information structure than but, which merely expresses contrast).

JASAN would tell us to use however wherever appropriate, but otherwise, and ALS would tell us to avoid but.  Either way, but would be disfavored.  In actual usage, but dwarfs however, though the frequency of however is not negligible, and is even considerable for some practiced writers (we exhibit some statistics).  We suggest that GR might reflect an appreciation of the discourse subordination of the material that however introduces, but we're dubious about GR on two grounds: other sentence-initial discourse connectives (consequently, therefore, nonetheless, nevertheless) that no one seems to complain about, although they are also discourse subordinators; and sentence-internal uses of however ("We at last succeeded, however, in reaching camp."), which are also discourse subordinators, but are often suggested as substitutes for initial however.  The other initial adverbials and the internal uses of however also share with initial however its prosodic weight and formality of style, so there seems to be no external justification for the bias against initial however, which remains a matter of individual taste -- perhaps a reaction to the over-use of initial however by student writers who have been taught NIC.  In any case, GR runs against both the Gricean JASAN and the prescriptivists' usual advice, ALS; there is no good reason not to use sentence-initial linking however on occasion.


Fraser, J. Bruce.  1998.  Contrastive discourse markers in English.  Jucker & Ziv 1998:301-26.

Garner, Bryan A.  1998. A dictionary of modern American usage.  NY: Oxford Univ. Press.

- 2003.  Garner's modern American usage.  [2nd ed. of Garner 1998]  NY: Oxford Univ. Press.

- 2004.  The winning brief: 100 tips for persuasive briefing in trial and appellate courts.  2nd ed.  NY: Oxford Univ. Press.

Jucker, Andreas H. & Yael Ziv (ed.).  1998.  Discourse markers: Descriptions and theory.  Amsterdam: Benjamins.

Schiffrin, Deborah.  1987.  Discourse markers.  Cambridge Univ. Press.

Strunk, William Jr.  1918.  The elements of style.  Ithaca NY: W.P. Humphrey. [Available on-line from 1999 here]

Strunk, William Jr. & E.B. White.  2000.  The elements of style.  4th ed. [1st ed. 1959]  NY: Longman.

Zwicky, Arnold M.  2006a. If they do it too much, they should be told not to do it at all.  (Language Log 10/31/06: link)

- 2006b. However,...  (Language Log 11/1/06: link)

zwicky at-sign csli period stanford period edu

Posted by Arnold Zwicky at February 21, 2007 12:42 AM